Research on the analysis and prediction of pavement crack causes based on particle swarm optimization neural network technology

The deterioration of asphalt highways in China, particularly due to the emergence of transverse cracks in semi-rigid base pavements, significantly undermines their structural integrity and overall performance. This study presents a comprehensive analysis conducted on a 3,000-kilometer highway network, utilizing a decade of operational, maintenance, and environmental data. The research applied Particle Swarm Optimization (PSO) and Back Propagation Neural Network (BPNN) algorithms to identify six critical factors influencing pavement crack formation: pavement layer information, layer thickness, total and surface layer service time, natural zoning, and traffic load. The findings reveal that these factors contribute differently to the occurrence and progression of pavement distresses under various conditions. Specifically, the study found that semi-arid regions exhibit a rapid increase in crack-related issues, thicker pavement layers slow down disease progression, and the extent of damage varies significantly across different service periods, with a notable acceleration observed in pavements with 4-12 years of service. Furthermore, a threshold effect was identified in the impact of traffic load, with a sharp increase in crack damage area when the Annual Average Daily Traffic (AADT) exceeds 5,000 vehicles per day. The predictive analysis using the PSO-BPNN model achieved an accuracy rate exceeding 90%, demonstrating its effectiveness in guiding maintenance decisions for highway pavements.